PIXEL-ORIENTED VISUALIZATION FOR EXPLAINING DATA CLASSIFICATION IN A MULTILAYER NEURAL
Palavras-chave:
Artificial Neural Network, SHAP, XAI, Information VisualizationResumo
Acompanhando o crescimnto de aplicativos que utilizam Inteligência Artificial, também crescem pesquisas recentes para explicar o funcionamento desses aplicativos e torná-los mais aceitáveis pelo homem. Este artigo apresenta uma explicação alternativa do processo de classificação de dados realizado por um algoritmo de Inteligência Artificial. Propomos uma abordagem de visualização de informação orientada a pixel para explicar o classificador perceptron multicamada usando SHAP. Observando os resultados obtidos, foi possível identificar as características relevantes para explicar a classificação.
Downloads
Referências
AHA, D. et al. UCI Machine Learning Repository. Available in: https://archive.ics.uci.edu/ml/about.html. Accessed: 2023 May 01.
ARRIETA, A. B. et al. Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities, and challenges toward responsible AI. Information Fusion, v. 58, p. 82-115, 2020. DOI: https://doi.org/10.1016/j.inffus.2019.12.012
BERG, S. et al. The fundamental package for scientific computing with Python. Available in: https://numpy.org/. Accessed: 2023 May 01.
BOISBERRANGER, J. et al. Machine Leaning in Python. Available in: https://scikit-learn.org/stable. Accessed: 2023 May 01.
HUNTER, J. D. Visualization with Python. Available in: https://matplotlib.org/. Accessed: 2023 May 01.
LUNDBERG, S. Welcome to the SHAP documentation. Available in: https://shap.readthedocs.io/en/latest/index.html. Accessed: 2023 May 01.
MALATO, G. How to explain neural networks using SHAP. Available in: https://www.yourdatateacher.com/2021/05/17/how-to-explain-neural-networks-using-shap/. Published: 2021 May 17. Accessed: 2023 May 01.
MARCÍLIO-JR, W. E.; ELER, D. M. Explaining dimensionality reduction results using Shapley values. Expert Systems with Applications, v. 178, p. 115020,
DOI: https://doi.org/10.1016/j.eswa.2021.115020
MAZZANTI, S. SHAP Values Explained Exactly How You Wished Someone Explained to You. Towards Data Science, 4 jan. 2020. Available in: https://towardsdatascience.com/shap-explained-the-way-i-wish-someone-explained-it-to-me-ab81cc69ef30. Accessed: 2023 May 01.
MILLER, T. Explanation in artificial intelligence: Insights from the social sciences. Artificial Intelligence, v. 267,pp. 1-38, 2019. DOI: https://doi.org/10.1016/j.artint.2018.07.007
PANTELIS, L.; PAPASTEFANOPOULOS, V.; KOTSIANTIS, S. Explainable AI: A Review of Machine Learning Interpretability Methods. Entropy , v.23, n.1, p., 18, 2021. DOI: https://doi.org/10.3390/e23010018
SAKURAI, R.; ZUCHI, J. D. As Revoluções Industriais até a Industria 4.0. Revista Interface Tecnológica, v. 15, n. 2, p. 480-491, 2018. DOI: https://doi.org/10.31510/infa.v15i2.386
WANG, P. et al. The World´s Must Popular Data Science Plataform. Available in: https://www.anaconda.com. Accessed: 2023 May 01.